Post on 25-Aug-2020
i
Medicare Graduate
Medical Education Funding
is Not Addressing the
Primary Care Shortage: We Need a Radically Different Approach
______________________________________________________
Bruce Steinwald, Paul Ginsburg, Caitlin, Brandt, Sobin Lee, and Kavita Patel
December 2018
This report is available online at: https://www.brookings.edu/research/medicare-graduate-medical-education-funding-is-not-addressing-the-primary-care-
shortage-we-need-a-radically-different-approach
USC-Brookings Schaeffer Initiative for Health Policy
i
Contents
Editor’s Note .......................................................................................................................................... ii
Statement of Independence ................................................................................................................... ii
Acknowledgments .................................................................................................................................. ii
Introduction ........................................................................................................................................... 1
The Structure and Financing of Medical Education in the US .............................................................. 2
The Demand For, And Supply Of, Medical Education Services ............................................................. 3
Specialty Mix and the Income Gap ........................................................................................................ 5
How Medicare Influences Medical Specialty Determination .............................................................. 10
Policy Implications ............................................................................................................................... 14
Conclusion............................................................................................................................................ 18
ii
EDITOR’S NOTE
This white paper is part of the USC-Brookings Schaeffer Initiative for Health Policy, which is a
partnership between the Center for Health Policy at Brookings and the USC Schaeffer Center for
Health Policy & Economics. The Initiative aims to inform the national health care debate with rigorous,
evidence-based analysis leading to practical recommendations using the collaborative strengths of
USC and Brookings.
STATEMENT OF INDEPENDENCE
The Brookings Institution is a nonprofit organization devoted to independent, in-depth research that
leads to pragmatic and innovative ideas on how to solve problems facing society. The conclusions and
recommendations of any Brookings publication are solely those of its author(s) and do not reflect the
views of the Institution, its management, or its other scholars.
The authors did not receive financial support from any firm or person for this article or from any firm
or person with a financial or political interest in this article. They are currently not an officer, director,
or board member of any organization with an interest in this article.
ACKNOWLEDGMENTS
We thank current and former officials of the Association of American Medical Colleges and the BJC
Health Care System for their opinions and insights on financing graduate medical education in the US.
We also thank Craig Lisk, Edward Salsberg, and Matthew Fiedler for their helpful comments on a
previous draft without associating them with any remaining errors.
1
Introduction
A growing body of evidence shows that areas with robust primary care systems tend to have better
outcomes and lower per capita costs than areas that rely more on specialists.1 Over the past several
decades, the US medical education system has produced an increasingly specialized physician
workforce without any strategic direction toward achieving a socially desirable mix of primary care
physicians (PCPs) and specialists.2 At the same time, health care reforms, such as patient-centered
medical homes and Accountable Care Organizations (ACOs), rely more on PCPs and other providers
who are equipped to coordinate their own care with the care of specialists. Demographic trends signal
a growing need for such coordination as the population ages and patients with multiple chronic
conditions become more prevalent.3 Despite these trends, physicians in training tend not to select
primary care or related specialties, making it difficult to achieve reforms that rely on an adequate
supply of PCPs.
In this paper, we examine the ways Medicare pays for physician and hospital services and subsidizes
graduate medical education (GME) in teaching hospitals.4 Looking at Medicare’s role through the lens
of the medical education marketplace, we conclude that Medicare’s GME subsidies have relatively little
effect on specialty mix. Medicare's physician payment policies, however, through their effect on
widening the gap between PCPs’ and specialists’ incomes, skew the choices made by doctors when
selecting residency positions and entering into medical practice and are therefore an important
determinant of the PCP/specialist mix in the US.
The next section provides an overview of how medical education is structured and financed in the US.
We then present an informal model of the market for GME services, followed by an analysis of how
the income gap between primary care and specialist doctors affects physician specialty mix in the US.
An examination of how Medicare influences the decisions of both consumers and suppliers of GME
services is followed by a discussion of policy options and a concluding section that summarizes
Medicare’s important role in medical specialty determination in the US.
1 Starfield, B., Shi, L., Macinko, J. (2005). Contribution of Primary Care to Health Systems and Health. Milbank Quarterly,
83(3), 457-502. https://www.ncbi.nlm.nih.gov/pubmed/16202000; Phillips, R. and Bazemore, A. (2010). Primary Care and
Why It Matters for U.S. Health System Reform. Health Affairs.
https://www.healthaffairs.org/doi/full/10.1377/hlthaff.2010.0020; Shi, L. (2012). The Impact of Primary Care: A Focused
Review. Scientifica, https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3820521/pdf/SCIENTIFICA2012-432892.pdf.
2 IOM (Institute of Medicine). 2014. Graduate medical education that meets the nation’s health needs. Washington, DC: The
National Academies Press. https://doi.org/10.17226/18754.
3 Dall, T. et al. (2013). An Aging Population and Growing Disease Burden Will Require A Large and Specialized Health Care
Workforce by 2025. Health Affairs. https://www.healthaffairs.org/doi/abs/10.1377/hlthaff.2013.0714.
4 The 4 years of medical school are referred to as “undergraduate medical education.” Graduate medical education (GME)
consists of residencies and fellowships served in teaching hospitals.
2
The Structure and Financing of Medical Education in the US
In 2016, 28,283 students enrolled in, and 22,477 graduated from, one of the 178 accredited allopathic
or osteopathic medical schools in the US.5,6 With very few exceptions, students matriculate for 4 years,
obtain an MD or DO degree, and then enter a residency program. Average tuition costs plus related
fees are substantial, ranging from a median cost of about $240,000 of attending a public medical
school for students graduating in 2016 or 2017 to a median cost of about $314,000 for attending a
private medical school. Despite the fact that about half of US medical students come from the
wealthiest fifth of households, more than 80 percent graduate with debt of $100,000 or more.7
Medical education continues when students go through a matching program that determines the
specialty and teaching hospital location of the residency.8 In 2018, 20,418 US allopathic and 5,770
osteopathic seniors and graduates entered a US residency program.9 These students were augmented
by both US and non-US International Medical Graduates (IMGs), raising the total number of entrants
into US residency programs by another 7,230.10 Residency programs vary by specialty and length, and
some students may complete more than one residency or enter a fellowship after completing a
residency. After at least one accredited residency, doctors may be certified in a specialty and enter into
medical practice.11
Financing of graduate medical education (GME) is a complex combination of revenues obtained by
teaching hospitals for the provision of health care services to patients plus additional payments made
in a variety of ways by payers for those services to subsidize the costs of training residents. These costs
include stipends paid to residents and residency program expenses including medical faculty salaries.
5 Allopathic and osteopathic physicians are both licensed to practice the full scope of medicine – osteopathic physicians tend to put
more emphasis on primary care, although these physicians can practice in any specialty. See more at AACOM,
https://www.aacom.org/become-a-doctor/us-coms, and AAMC https://www.aamc.org/about/. Note: there are now 185 accredited
schools.
6 Dalen, J., Ryan, K. (2016). United States Medical School Expansion: Impact on Primary Care. American Journal of Medicine, Vol
129(1). https://www.amjmed.com/article/S0002-9343(16)30554-X/pdf.
7 Marcu, et al. (2017). Borrow or Serve? An Economic Analysis of Options for Financing a Medical School Education. Academic
Medicine, 92(7); 966-97. https://www.ncbi.nlm.nih.gov/pubmed/28121649.
8 Learn more about the matching program at http://www.nrmp.org/about-nrmp/.
9 As the system moves to a single match, most osteopathic graduates now match through the NRMP match, while the rest match
through the traditional DO AOA Match. NRMP 2018 Residency Match Results, https://mk0nrmpcikgb8jxyd19h.kinstacdn.com/wp-
content/uploads/2018/04/Main-Match-Result-and-Data-2018.pdf; AOA 2018 Residency Match Results,
https://natmatch.com/aoairp/stats/2018prgstats.html.
10 NRMP. (2018) Main Resident Match by the Numbers. http://www.nrmp.org/wp-content/uploads/2018/03/2018-Match-by-the-
Numbers.pdf.
11 Association of American Medical Colleges (AAMC). The Road to Becoming a Doctor,
https://www.aamc.org/download/68806/data/road-doctor.pdf.
3
The subsidies may be explicit, such as Medicare’s subsidies (which will be discussed later), or implicitly
included in the health service payments negotiated between private insurers and teaching hospitals.
Differing views of the principal role of the resident in the teaching hospital are somewhat controversial,
ranging from; essentially, an advanced student who participates in the care of patients as part of the
teaching curriculum to a valuable professional whose contribution to teaching hospital and faculty
practice revenues exceeds the costs of his or her training. Without attempting to resolve this issue, we
focus in the next section on how specialty training’s effect on practice income is a determinant of
specialty choice.
The Demand For, And Supply Of, Medical Education Services
Because the number of medical school positions is limited, the application process is very competitive.
Undergraduate baccalaureate performance, score on the Medical College Admission Test (MCAT),
relevant experience, and other factors determine which applicants are accepted.12 Once in medical
school, the competition shifts to application for desirable residency positions. Performance on medical
school exams and general exams, such as the United States Licensing Medical Exam, Step 1,
traditionally taken during the second of the four medical school years, relevant experience and other
factors determine which graduating students are able to land their preferred residencies in their
preferred locations.13 Taking these preferences into account, and those of the teaching hospitals, the
National Resident Matching Program (NRMP) uses a computer algorithm to match students with
programs in order to maximize their collective preferences. While not all students who attend school
in the United States obtain their top residency choices, nearly all are able to enter a US residency
program, with PGY-1 match rate for allopathic graduates being 92-95 percent historically.14
Residency may be in primary care or non-primary care specialties. For reasons discussed later, US
medical school graduates tend to prefer non-primary care specialties,15 so a relatively high proportion
of primary care residency positions are filled by IMGs from a variety of countries and by US citizens
studying abroad, often in “off-shore” medical schools.16 Counting the four medical school years, the
12 Ibid.
13 Green M. (2009). Selection Criteria for Residency: Results of a National Program Directors Survey. Academic Medicine,
84(3):362-7. https://www.ncbi.nlm.nih.gov/pubmed/19240447
14 NRMP. (2017). Press Release: 2017 NRMP Main Residency Match the Largest Match on Record. http://www.nrmp.org/press-
release-2017-nrmp-main-residency-match-the-largest-match-on-record/
15 The number of students entering primary care residencies, particularly internal medicine, is actually overstated because many
intend to specialize once they complete the initial residency.
16 National Resident Matching Program. (2018). Results and Data: 2018 Main Residency Match. National Resident Matching
Program. http://www.nrmp.org/wp-content/uploads/2018/04/Main-Match-Result-and-Data-2018.pdf; GAO (2009). GAO Report:
Graduate Medical Education: Trends in Training and Student Debt. https://www.gao.gov/assets/100/96096.pdf . Several
organizations representing primary care physicians have recently expressed concern that changing US policy making IMG entry into
4
total time in training ranges from a minimum of seven to eleven years (or even longer in some cases)
before the doctor becomes board certified in a specialty.17 It is a long and arduous process but medicine
continues to be an attractive profession and a sound investment.18
Graduating medical students’ preferences for residencies in different specialties reflect a number of
pecuniary and nonpecuniary factors. Survey data suggest that the match between specialty content
and individual interests and skills, achieving a favorable work/life balance, and the influence of
mentors rank among the most important stated reasons for selecting a residency in a given specialty.19
As we discuss later in our section on the income gap, however, we believe that the evidence shows that
eventual earnings are a very important determinant of specialty choice.
The supply side in the market for medical education is complicated because educational services and
health care services are jointly produced in teaching hospitals and residents are both consumers of
educational services and producers (in conjunction with medical school faculty and staff) of health
care services.20 Thus, the way in which Medicare and other payers pay for hospital and physician
services performed in teaching hospitals may be a factor in determining what residencies and
residency positions are offered to graduating medical students. That is, the contribution of residents
to hospital and medical faculty practice plan revenues may vary by specialty and the mix of primary
care and non-primary care residents may influence the extent to which residents’ outputs contribute
to hospital and teaching physicians’ revenues.21
Because of the joint production phenomenon, teaching hospitals’ decisions about which residencies to
support, and how many positions in each to offer, is not as simple as responding only to graduating
medical students’ demands. Decision makers need to balance the costs of training residents in
particular residencies against the revenues that residents generate working with teaching physicians
and hospital staff.22 These revenues may devolve to the benefit of the hospital through its various
US residencies more difficult could disrupt teaching hospitals’ provision of care. See letter to L Francis Cissna, Director, US
Citizenship and Immigration Services, dated May 30, 2018.
17 AAMC. The Road to Becoming a Doctor. https://www.aamc.org/download/68806/data/road-doctor.pdf.
18 Roth, Nicholas. The Costs and Returns to Medical Education.
https://www.econ.berkeley.edu/sites/default/files/roth_nicholas.pdf
19 AAMC. (2017). Graduation Questionnaire. https://www.aamc.org/data/gq/.
20 MedPAC. (2010). Report to Congress: Aligning Incentives in Medicare. http://www.medpac.gov/docs/default-
source/reports/Jun10_EntireReport.pdf
21 Wynn, B.O., Smalley, R., Cordasco, K.M. (2013).Does it Cost More to Train Residents or to Replace Them? A Look at the Costs and
Benefits of Operating Graduate Medical Education Programs. RAND Corporation.
https://www.rand.org/pubs/research_reports/RR324.html
22 IOM (Institute of Medicine). 2014. Graduate medical education that meets the nation’s health needs. Washington, DC: The
National Academies Press. https://doi.org/10.17226/18754; Council on Graduate Medical Education: Twentieth Report. (2010)
Advancing Primary Care. https://www.hrsa.gov/advisorycommittees/bhpradvisory/cogme/Reports/twentiethreport.pdf; MedPAC
5
inpatient and outpatient services or to the medical staff through performance of physician and related
services, and there may be a variety of arrangements for revenue sharing, including with the larger
entity (such as a university) that the hospital and medical staff are a part of.
Just as medical students have different preferences and capabilities affecting specialty choice, teaching
hospitals display substantial differences in their preferences for residency programs. In addition to
revenue concerns, other factors such as prestige, resident work contribution, and
intellectual/academic opportunities also may be influential determinants of residency offerings.23 In
addition, certain teaching hospitals, particularly the smaller ones, appear to have more success in
“producing” physicians in primary care specialties that their larger and more urban counterparts.24
Specialty Mix and the Income Gap
In 2016, the proportion of practicing physicians in the US was approximately 68 percent specialists
and 32 percent PCPs.25 The Council on Graduate Medical Education (COGME), an organization that
periodically reviews the composition and performance of the physician workforce, recommends that
the ideal proportion of generalists should be at least 40 percent.26 Kaiser-Permanente, the large
integrated delivery system that employs its physicians, maintains its PCP proportion at 45 percent.27
Several authorities, including the Association of American Medical Colleges and the Institute of
(2010). Report to the Congress: Medicare Payment Policy. http://www.medpac.gov/docs/default-
source/reports/Mar10_EntireReport.pdf COGME(2010) ; MedPAC(2010)
23 Sivey, P., Scott, A., Joyce, C., Humphreys, J. (2012). Junior Doctors’ Preferences for Specialty Choice. J. Health Econ. 2012 Dec;
31(6):813-23. https://www.ncbi.nlm.nih.gov/pubmed/22940638
24 Chen,C., Petterson, S., Phillips, R., et al. (2013). Toward Graduate Medical Education (GME) Accountability: Measuring the
Outcomes of GME Institutions. Academic Medicine. September 2013. Volume 88 (9):1267-80.
https://journals.lww.com/academicmedicine/Fulltext/2013/09000/Toward_Graduate_Medical_Education__GME_.31.aspx
25 AAMC (2018). The Complexities of Physician Supply and Demand: Projections from 2016 to 2030. https://aamc-
black.global.ssl.fastly.net/production/media/filer_public/85/d7/85d7b689-f417-4ef0-97fb-
ecc129836829/aamc_2018_workforce_projections_update_april_11_2018.pdf. Note: The 32% includes primary-care trained
hospitalists, which make up 4% of the workforce.
26 Council on Graduate Medical Education: Twentieth Report. (2010) Advancing Primary Care.
https://www.hrsa.gov/advisorycommittees/bhpradvisory/cogme/Reports/twentiethreport.pdf;. The issue of specialist/PCP
imbalance is not confined to the US. The World Health Organization (WHO), noting that the ratio of PCPs to specialist in the
European Union (EU) is 1:3.2, recently issued a call for increasing PCP slots in the EU. See http://www.euro,who.int/health-
topics/Health-systems/health-workforce/data-and-statistics
27 Reinke, T. (2009). Are health plans responding to primary care shortage? Managed Care Magazine.
https://www.managedcaremag.com/archives/2009/2/are-health-plans-responding-primary-care-shortage. The Kaiser Permanente
proportion may be a bit overstated owing to the fact that some specialty services furnished to its beneficiaries are outsourced to
specialists in the community.
6
Medicine, have asserted that the PCP/specialist imbalance is due to a shortage of PCPs in the US.28
Some analysts have suggested that nurse practitioners (NPs) and Physician Assistants (PAs), who
currently supply approximately one-fourth of primary care services in the US, may be increasingly
relied upon in areas where PCPs are in short supply.29 Recent evidence indicates, however, that NPs
and PAs have a tendency to devote their services increasingly in specialty, as opposed to primary care,
practices.30 While it is true that the growing number of PAs and NPs relieves some of the pressure of
underproducing PCPs, because of state scope of practice laws and intrinsic limitations, the extent to
which PAs and NPs can substitute for PCPs is uncertain.31 Many NPs and PAs are compensated under
the same Medicare fee schedule as physicians and are subject to the same incentives facing physicians
making specialty decisions discussed in the next section.
A potentially important element in the demand for education in a primary care versus a non-primary
care specialty is the so-called “income gap,” that is, the extent to which the income of MDs in many
28 AAMC. 2018 Update: The Complexities of Physician Supply and Demand: Projections from 2016 to 2030. https://aamc-
black.global.ssl.fastly.net/production/media/filer_public/85/d7/85d7b689-f417-4ef0-97fb-
ecc129836829/aamc_2018_workforce_projections_update_april_11_2018.pdf
29IOM (Institute of Medicine). 2014. Graduate medical education that meets the nation’s health needs. Washington, DC: The
National Academies Press. https://doi.org/10.17226/18754.
30 Medical Bag. (2014). More PAs Choosing Specialties over Primary Care. https://www.medicalbag.com/physician-assistants/more-
pas-choosing-specialties-over-primary-care/article/472697/ ; Coombs, L. (2015) The Growing Nurse Practitioner Workforce in
Specialty Care. Journal for Nurse Practitioners, Vol 11(9):907-909
https://www.sciencedirect.com/science/article/pii/S1555415515007096
31 See Salsberg (2018) for a discussion of how the increasing supply of NPs and PAs is expanding access to primary care services.
7
specialties exceeds that of a PCP. In this sense, the demand for residency positions is derived, in part,
from the eventual income earned after completing education and entering into medical practice. In
the US, average compensation does vary substantially by specialty, as indicated in Exhibit 1. The
primary care specialties of Pediatrics, Family Medicine, and Internal Medicine tend to be at the low
end of the compensation distribution.32 That there is an income gap is not surprising because specialty
residencies (as well as fellowship training for subspecialties, such as cardiology) tend to be longer with
mastery of additional procedure-based skills required as part of successful training. A typical primary
care residency is approximately three years in duration while a typical specialty residence, such as
orthopedic surgery, approximately five years. There are gaps in earnings across the spectrum of
specialties and, as stated above, there are many other factors influencing specialty choice. However,
the primary care vs specialist gap in the US seems exceptionally high in relation to the extra time
invested; the magnitude of the gap is not seen elsewhere, for instance, in most European countries.33
Fifteen years ago, Reinhardt and Anderson showed that a sufficient explanation for why health care
spending per capita is so much higher in the US than in most of Europe is because we pay our providers
and manufacturers – hospitals, doctors, drug companies, etc. – much more in the US than elsewhere.34
OECD data indicate that this trend persists. Not only is the gap between PCPs and specialists higher
in the US, it is mainly because US specialists’ compensation exceeds other countries’ specialist
compensation more than US PCP compensation exceeds other countries’ PCP compensation. With few
exceptions, the proportion of physicians who are PCPs is much lower in the US than in other
countries.35, 36
But how important is the income gap to this process compared to all the other factors that could affect
specialty preferences outlined in the previous section? Some studies have shown that expectation of
future income is an important determinant of students’ choice of specialty, as opposed to primary care,
32 One study estimated that a PCP would have to receive a bonus of $1.1 million upon completion of medical school for that PCP’s
lifetime earnings to equal those of a cardiologist. See Vaughn et al. (2010). https://aspe.hhs.gov/report/health-practitioner-
bonuses-and-their-impact-availability-and-utilization-primary-care-services/1-effects-earnings-specialty-choice-physicians
33 Laugesen, M.J., Glied, S.A. (2011). Higher Fees Paid to US Physicians Drive Higher Spending for Physician Services Compared to
Other Countries. Health Affairs, Vol 30(9). https://www.healthaffairs.org/doi/10.1377/hlthaff.2010.0204; Conover, Chris. (2013).
Are U.S. Doctors Paid too Much? Forbes. https://www.forbes.com/sites/theapothecary/2013/05/28/are-u-s-doctors-paid-too-
much/#3c8df023d525 (See Table 2)
34 Anderson, G., Reinhardt, U., Hussey, P., Petrosyan, V. (2003). It’s the Prices, Stupid: Why the United States is so Different from
Other Countries. Health Affairs Vol 22(3). https://www.healthaffairs.org/doi/pdf/10.1377/hlthaff.22.3.89 OECD (2017). Health at
Glance 2017: OECD Indicators. https://www.health.gov.il/publicationsfiles/healthataglance2017.pdf
35 OECD (2017). Health at a Glance 2017: OECD Indicators. https://www.health.gov.il/publicationsfiles/healthataglance2017.pdf
36 Recent survey data from the Medical Group Management Association show that the cumulative increases in PCP and specialist
compensation in the US were about 10 and 5 percent, respectively, over the past 5 years. (MGMA press release, May 16, 2018.) While
this is an encouraging trend, it is insufficient to reduce the dollar income gap between PCPs and specialists.
8
residencies.37 This evidence includes the propensity of residents in Internal Medicine to pursue
subspecialty fellowships, thereby moving out of primary care into specialty practice.38 In 2014, the
Office of the Assistant Secretary of Health and Human Services reviewed five studies presenting
relevant evidence on the matter.39 Except for one study that attempted to estimate the effect of
consultation fees on propensity to enter general practice in Canada, the implied elasticities of the other
studies, which focused on earnings differences, were quite large. That is, according to these studies, a
one percentage point reduction in the income gap between PCP and specialist physicians would lead,
as a crude approximation, to about a one percentage point, or more, increase in the PCP proportion.
37 Ebell, Mark H. (1989). Choice of Specialty: It’s Money that Matters in the USA. JAMA. 1989; 262(12):1630.
https://jamanetwork.com/journals/jama/article-abstract/1725027 (1989), Ebell, Mark H. (2008). Future Salary and US Residency
Fill Rate Revisited. JAMA 2008 Sep 10:300(10):1131-2. https://jamanetwork.com/journals/jama/article-abstract/182526
38 Halvorsen, Andrew J., Kolars, Joseph C., McDonald, Furman S. (2010) Gender and Future Salary: Disparate Trends in Internal
Medicine Residents. Am J Med. Vol 123(5):470-75. https://www.amjmed.com/article/S0002-9343(10)00063-X/abstract
39 ASPE. (2014). Health Practitioner Bonuses and their Impact on the Availability and Utilization of Primary Care Services. 1. Effects
of Earnings on Specialty Choice by Physicians. https://aspe.hhs.gov/report/health-practitioner-bonuses-and-their-impact-
availability-and-utilization-primary-care-services/1-effects-earnings-specialty-choice-physicians
9
Exhibit 2 illustrates the pathways of decision-making at various stages of medical training. By the time
medical students approach graduation, they are aware of the income gap and the relationship between
residency specialty and practice specialty, and express their preferences through the residency
matching program. These preferences are further expressed when residents enter specialty and sub-
specialty residencies and fellowships.
10
That students’ preferences are different for specialist and generalist residencies can be inferred from
the proportions of resident positions filled by graduates of US allopathic medical schools. When
considering alternative specialty choices, medical students are sometimes advised to follow the
EROAD.40 That is, the specialties of Emergency Medicine, Radiology, Ophthalmology, Anesthesiology,
and Dermatology are singled out as specialties that offer both substantial income and favorable
lifestyle balance that many students would find personally rewarding, especially compared to primary
care specialties, when entering medical practice.41 Exhibit 3 contrasts the relatively high US medical
school graduate fill rates for the EROAD specialties with the lower fill rates for primary care
residencies, particularly in family medicine and internal medicine.42 It’s also important to note that
approximately 81 percent of internal medicine residents will subspecialize, along with 41 percent of
pediatric residents, decreasing the number of residents who enter into primary care practice.43
It's important to recognize that the income gap and its effect on specialty choice is a demand-side
phenomenon. While this may seem self-evident, a number of policy solutions to the specialty
imbalance, particularly in the context of Medicare payment policy, focus on supply-side issues of
financing medical education in teaching hospitals. However, there may be ways to make the incentives
of doctors in training and teaching hospitals more closely aligned, as suggested in the following
sections.
How Medicare Influences Medical Specialty Determination
After Congress changed Medicare’s inpatient hospital reimbursement from a cost-based system to a
Prospective Payment System (PPS) of paying for Diagnosis-Related Groups (DRGs) in 1983, a unique
system of subsidizing the costs of training residents in teaching hospitals was added to the payment
formula.44 The formula called for Medicare to pay its share of the direct costs of Graduate Medical
Education (DGME), such as residents’ stipends and teaching physicians’ salaries, and a portion of the
40 Weida, Phillips, Bazemore et al. (2010). Loss of Primary Care Residency Positions Amidst Growth in Other Specialties. Robert
Graham Center. https://www.graham-center.org/rgc/publications-reports/publications/one-pagers/loss-of-primary-care-
residencies-2010.html
41 Blair, Jenny. Taking the E-ROAD. Yale Medicine Magazine. http://ymm.yale.edu/autumn2007/features/feature/51534/ ;
Rayburn, William F. & Schulkin, Jay. (2011). Changing Landscape of Academic Women’s Health Care in the United States. Springer.
https://books.google.com/books?id=oO4eQte2zG0C&pg=PA34&lpg=PA34&dq=e-
road+specialties&source=bl&ots=hjKhQR9PzF&sig=Nl6sXUoRZJmiE04uOlMZCgVb71k&hl=en&sa=X&ved=0ahUKEwiKvpejnt3Z
AhWpneAKHdmEAKw4ChDoAQhXMAk#v=onepage&q=e-road%20specialties&f=false
42 More broadly, the correlation between average practice compensation in 14 specialties and the percent of residency positions in
those specialties filled by US medical school graduates is strongly positive (R=0.71, p < .01).
43 Dalen, J., Ryan, K. (2016). United States Medical School Expansion: Impact on Primary Care. American Journal of Medicine, V129(1). https://www.amjmed.com/article/S0002-9343(16)30554-X/pdf.
44 The Medicare GME subsidy was initiated in 1985. See https://www.cms.gov/Medicare/Medicare-Fee-for-Service-
Payment/AcuteInpatientPPS/DGME.html
11
Indirect Cost of Graduate Medical Education (IME), such as higher inpatient costs owing to residents
ordering extra tests and generally reducing the efficiency of hospital operations.45
In 1992, Congress changed the method of reimbursing physicians for their services to Medicare
patients from a system based on individual physicians’ historical charges to one based on the “relative
values” of the thousands of services physicians charge Medicare for – the Resource-Based Relative
Value Scale (RBRVS) fee schedule. As discussed below, the ways in which Medicare fees for the services
performed by PGPs and specialists are determined and updated are an important determinant of inter-
specialty differences in incomes.
Medicare influences the income gap primarily through its physician fee schedule (PFS). For many
reasons, which have been written about extensively, the process of updating the fee schedule (which
consists of modifying existing fees and establishing fees for new procedures) has benefitted procedure-
oriented specialties at the expense of cognitive specialties.46 Fees for existing services have increased
very little over the past several years, but high payments for new services coupled with substantial
increases in the volume of expensive diagnostic and other procedures, have contributed to Medicare’s
share of the income gap.47 Physicians in procedure-oriented specialties have opportunities to harness
changes in medical technology to increase their incomes while primary care physicians and others in
cognitive specialties are constrained by the fact that their services are largely a function of, and
therefore limited by, their own time.48 Other payers do not necessarily pay the same rates as Medicare
for the same services, but they tend to use the Medicare RVS as a starting point, which preserves the
substantial difference in payments between procedural and cognitive services.49
If the income gap is a product of the way that Medicare and other payers reimburse physicians for their
services, what is the role of teaching hospitals in the determination of specialty mix? Numerous studies
have concluded that Medicare’s subsidization of medical education, particularly the IME add-on to
inpatient DRG payments in teaching hospitals, might be an important factor. For example, in 2012 the
45 Heisler, E.J., Janse, D.J., Mitchell, A., Viranga Panangala, S., Talaga, S.R. (2016). Federal Support for Graduate Medical
Education: An Overview. Congressional Research Service. https://fas.org/sgp/crs/misc/R44376.pdf.
46 GAO (2015). Report to Congressional Committees. Medicare Physician Payment Rates: Better Data and Greater Transparency
Could Improve Accuracy. https://www.gao.gov/assets/680/670366.pdf
47 USC-Brookings Schaeffer Initiative for Health Policy (2017). The Medicare Physician Fee Schedule Likely to Serve as Foundation
for Alternative Payment Models. The Brookings Institution. https://www.brookings.edu/wp-content/uploads/2017/08/medicare-
pfs-conference-brief-event-summary.pdf
48 Steinwald AB. (2008). Primary care professionals: recent supply trends, projections, and valuation of services. Testimony before
the Committee on Health, Education, Labor and Pensions, U.S. Senate (GAO-08-472T). Washington, DC: Government
Accountability Office, February 12, 2008. https://www.gao.gov/new.items/d08472t.pdf?source=ra
49 Maeda, J. L, Nelson, L. (2017). An Analysis of Hospital Prices for Commercial and Medicare Advantage Plans. Presentation.
Congressional Budget Office (CBO). https://www.cbo.gov/system/files/115th-congress-2017-2018/presentation/52819-
presentation.pdf
12
Institute of Medicine (IOM) convened a blue ribbon panel, chaired by two former Medicare program
administrators, to examine GME financing and recommend changes in the public interest.50 In its
2014 report, the IOM panel identified a mismatch (i.e., an overproduction of specialist and an
underproduction of generalist physicians) between the health needs of the population and specialty
makeup of the physician workforce as one of the deficiencies of Medicare’s GME financing.51 The
report recommended several significant changes to Medicare’s GME financing system but it did not
identify the Medicare PFS contribution to the income gap as a key factor in the PCP/specialist
imbalance.
The trend in residency positions and composition over the years indicates residents’ valuable
contributions to teaching hospitals’ revenues. Through the Balanced Budget Act of 1997 (BBA),
Congress placed a cap on Medicare GME funding and various stakeholders, including the Association
of American Medical Colleges (AAMC) which represents teaching hospitals, have argued that the cap
should be lifted in order to assure an adequate supply of physicians to meet the needs of an aging
population. The IOM report, however, noted that an increase in GME funding was unnecessary to the
production of more physicians.52 Enrollment in US medical schools during the current millennium
increased considerably despite the fact that most of the new residency positions over this period were
subject to the cap and did not generate additional Medicare GME revenues. Moreover, the specialist
residency FTEs increased by a larger percentage than PCP residency FTEs over the post-BBA period.53
Of course, Medicare is not the only source of financing GME. Medicaid, other HHS programs, the
Veterans Administration (VA), and private insurers pay as well, although it is often difficult to
determine how much. But, in terms of level of expenditure, Medicare dominates other payers’
influences. As indicated in Exhibit 5, for example, the annual Medicare spending on DGME and IME
exceeds $11.3 billion, far more than any other targeted subsidies to channel medical education in
socially desirable directions. MedPAC analysis has shown that the IME adjustment is approximately
twice what can be justified based on the costs of training residents.54
As suggested in our discussion of the market for medical education services, because residents support
the provision of services in their specialty areas, teaching hospitals must consider the effects of
50 IOM (Institute of Medicine). 2014. Graduate medical education that meets the nation’s health needs. Washington, DC: The
National Academies Press. https://doi.org/10.17226/18754.
51 The report states, “Hospitals’ control over the allocation of GME funds may also encourage the overproduction of specialists in
disciplines that generate financial benefits for an individual institution rather than for the health care system overall.” P. 101
52 IOM (Institute of Medicine). 2014. Graduate medical education that meets the nation’s health needs. Washington, DC: The
National Academies Press. https://doi.org/10.17226/18754.
53 At various times the US Congress enacted additional adjustments to Medicare GME payments to encourage primary care but, like
the GME payment cap, they did not materially affect the specialist/PCP imbalance.
54 Miller, M. E. (2015). Hospital Policy Issues. MedPAC. http://www.medpac.gov/docs/default-source/congressional-
testimony/testimony-hospital-policy-issues-ways-and-means-.pdf?sfvrsn=0
13
residency offerings on the profitability or unprofitability of hospital care to Medicare patients. In
general, surgical cases tend to be more profitable than nonsurgical cases,55 and the IME adjustment
tends to exacerbate this phenomenon. For example, the IME add-on for a relatively less expensive
case, such as pneumonia, might add about $30 to the DRG payment, while the add-on for a more
complex case, such as bypass surgery, might add ten times as much to the DRG payment. In addition,
there are no add-on payments for care provided in the outpatient setting where much of the work of
primary care residents and PCPs is conducted.
Because of the amount of money involved and the lack of accountability on how teaching hospitals
spend it, it is not surprising that the IOM study, MedPAC, and other authorities have suggested that
Medicare should reprogram the DGME and IME subsidies in the public interest. While such proposals
may have merit, it is hard to understand how they might reverse the incentives associated with the
income gap. Moreover, while the way that Medicare subsidizes medical education in teaching hospitals
is not a determinant of the income gap, it does appear that teaching hospitals may have an interest in
maintaining the status quo in preserving the specialist-oriented mix of residency positions.56
55 Koch, E., Becker, S. (2004). MedPAC Study – A Review of Findings Released October 28-29, 2004. MedPAC.
https://www.beckershospitalreview.com/pdfs/articles/MedPac.pdf; Shi, L., Sing, D.A. (2011) The Nation’s Health. Jones & Bartlett
Learning.
https://books.google.com/books?id=ttJIwRGuDPcC&pg=PA558&lpg=PA558&dq=are+some+DRGS+more+profitable+than+other
s+specialties+medicare&source=bl&ots=L3ViZcrC2I&sig=e_-
%20XbIrXIFLmf5Cz5BpP2xlFjeA&hl=en&sa=X&ved=0ahUKEwj79JDxscvZAhXjUN8KHXRmCfUQ6AEIKzAA#v=onepage&q=are
%20some%20DRGS%20more%20profitable%20than%20others%20specialties%20medicare&f=false
56 Weida et al. (2010). Loss of Primary Care Residency Positions Amidst Growth in Other Specialties. Robert Graham Center.
https://www.graham-center.org/rgc/publications-reports/publications/one-pagers/loss-of-primary-care-residencies-2010.html
14
Policy Implications
Our analysis indicates that, if we want to raise the proportion of doctors who are PCPs in the US, we
can’t afford to ignore the income gap. In light of the apparent oversupply of specialists relative to PCPs
in the US, it may seem surprising that the prices paid to specialists remain so stubbornly high. The
reason that the market doesn’t reduce the gap organically is due to the artificial way that different
services are priced in the US, particularly through the Medicare relative value scale and its adoption
by other payers. Raising prices for services that PCPs perform without lowering prices for tests,
imaging, and services that specialists perform would reduce the gap but increase Medicare spending
substantially. Lowering the prices of services that specialists perform would also reduce the gap, but
the history of Medicare payment policy shows that payment reductions, however, sensible, are difficult
to achieve. Nonetheless, Medicare’s contribution to the gap, and the harm that results from it, would
suggest that fundamental changes are warranted.
At various times, the US Congress enacted adjustments to Medicare GME payments to encourage
primary care. For example, Congress placed a two-year freeze on increases in the per-resident subsidy
for non-primary care residents, and applied a weighting factor to adjust direct GME payments in favor
of primary care residents. None of these changes had a substantial effect on Medicare payments and,
like the GME payment cap discussed above, they did not materially affect the specialist/PCP
imbalance.57
Since Medicare’s contribution to the gap operates through the PFS, less reliance on the PFS to
compensate PCPs might narrow the gap. In 2015, for example, MedPAC recommended augmenting
PCP compensation with per-beneficiary payments to supplement per-service payments.58 More
recently MedPAC has explored changes to the fee schedule, such as raising the fees for evaluation and
management services provided in the outpatient setting, services that furnish the bulk of Medicare
income to PCPs, and reducing payment for other services such that the net effect is budget-neutral.
Not only would this reduce the PCP/specialist income gap somewhat, it would also benefit physicians
in cognitive specialties who, like PCPs, rely on patient visits to generate Medicare income.59
57 See https://www.aamc.org/advocacy/gme/71152/gme_gme0001.html for an explanation of Medicare GME subsidy payments and
adjustments over time.
58 MedPAC (2015). Report to the Congress: Medicare Payment Policy. http://www.medpac.gov/docs/default-
source/reports/mar2015_entirereport_revised.pdf. While we focus here on the PFS, growth of Alternative Payment Models (APMs)
is another important mechanism for disconnecting PCP compensation from the per-service PCP/specialist payment differences
driven by the RBRVS.
59 Winter, A., Hayes, K. (2018). Medicare accountable care organization models: Recent performance and long-term issues.
MedPAC. http://www.medpac.gov/docs/default-source/default-document-library/rebalancing-fee-schedule-towards-e-m_jan-
2018_public.pdf?sfvrsn=0
15
Thus, our principal recommendation for Medicare to reduce the income gap is to alter relative
payments under the PFS, but what about GME financing policy? Are there any changes there that
could also favorably influence the PCP/specialist mix? Our analysis would suggest that just focusing
on the supply side, such as forcing teaching hospitals to increase PCP residencies in order to maintain
their GME subsidies, would be unlikely to result in more PCPs in practice. However, there may be
policies that focus directly on physicians in training that might have some effect on specialty choice.
We should recognize that the income stream of a physician begins with four years of expense and
foregone earnings during medical school, followed by three or more years of modest earnings during
residency and fellowship,60 before the physician enters practice and begins to realize a return on
investment in medical education. Because of the time value of money and difficulties and expense of
long-term borrowing by these individuals, policies that address a dollar of relative expense and income
during these early years should have a disproportionate effect on specialty choice compared to a dollar
of income years later in medical practice. Subsidies that are paid directly to doctors in training rather
than to teaching hospitals would reduce the effect of the income gap and therefore be a potential tool
to modify the effect of the income gap.
In a 2016 report, the Congressional Research Service summarized 25 federal loan repayment and
forgiveness programs for health care and public health professions.61 The most well-known of these
are National Health Service Corps programs operated by the Health Resources and Services
Administration, one of which is designed to encourage primary care physicians to practice in health
professional shortage areas. However, no programs exist solely to encourage doctors in training to
select primary care practice over specialization without geographical strings attached. A loan
forgiveness program that subsidizes medical school tuition and expenses in return for some minimum
number of years practicing in a primary care specialty (say, one year of practice for every year of
subsidy) could have a considerable effect on specialty choice.
In a 2017 article, Marcu, et al. provided data and analysis suitable for estimating the effect of a loan
forgiveness program in reducing the income gap between PCPs and specialists.62 Forgiving loans
totaling $180,000 (the median level of debt for students entering medical school in 2013) for the 4
years of medical school, in return for entering and remaining in a primary care specialty, would have
the effect of substantially reducing the net present value (NPV) difference in income streams of PCPs
compared to specialists who would borrow the same amount but repay the loans over a 10-year period
60 Resident and fellow salaries are determined largely by their year of training rather than the specialty they are in.
61 Hegji, A., Smole, D.P., Heisler, E. (2016). Federal Student Loan Forgiveness and Loan Repayment Programs. Congressional
Research Service. https://fas.org/sgp/crs/misc/R43571.pdf
62 Marcu, et al. (2017). Borrow or Serve? An Economic Analysis of Options for Financing a Medical School Education. Academic
Medicine, 92(7); 966-97. https://www.ncbi.nlm.nih.gov/pubmed/28121649.
16
after residency.63 For example, this subsidy would raise the NPV of going into General Internal
Medicine in Washington, DC, from $1,456,000 (total loan repayment) to $1,835,000 (no loan
repayment), a NPV that would exceed that of the ophthalmology NPV (total loan repayment) of
$1,573,000.64
Of course, a subsidy program of this nature would be expensive, in large part because students who
would enter primary care without financial incentives would receive the same subsidy as those who
were induced to enter primary care by the loan forgiveness program. As a crude approximation, if all
students entering family medicine or general internal medicine residencies took the loan with the
intention of having it forgiven, the annual cost would be about $2 billion.
This cost could be reduced if the amount of the subsidy were linked to medical school performance,
such as score on the Step 1 exam and other performance metrics. In addition, making the subsidy
performance based would encourage high-performing students to enter primary care and reduce the
free-rider problem. It could be especially important to encourage high performing students to enter
primary care practice as many trends in payment reform are relying on PCPs to “captain the ship” in
team-based medicine where many physician and non-physician providers are members of the care
team. In addition, this approach might also reduce the negative impression that some students have
that primary care is a refuge for low-performing or unambitious students.65
Another mechanism to reduce such debt burden is to eliminate tuition for medical students; most
recently New York University (NYU) School of Medicine announced that they would provide grants to
cover tuition for all of their medical students, regardless of need.66 Tuition burden reduction could
provide an upstream incentive of some significance that could influence specialty choice, but the
income gap may be too powerful to overcome through just the elimination of medical school tuition.
Subsidizing all students’ educational expenses would actually preserve the income gap differences
between PCPs and specialists. If one wants to encourage students to elect primary care, it would be
much more effective and efficient to subsidize only PCPs’ educational costs.
We acknowledge that there are many programs in effect to raise access to primary care services. It is
beyond the scope of this analysis to evaluate these programs individually but we note that the amount
of funding of such programs is small relative to the magnitude of the specialty mix problem and to the
63 Marcu et al’s analysis includes a Federal Stafford loan interest rate of 6.21 percent and current estimates of resident salaries and
annual incomes in various specialties over a 30-year career. Marcu et al did not analyze a full subsidy loan program but they did
analyze a no-strings scholarship program, which we regard as computationally equivalent to a full subsidy loan program.
64 Ophthalmology was selected as an example because it is one of the EROAD specialties; however, the gap would remain large in
some specialties, such as orthopedic surgery.
65 Insight provided by 3d year medical student at Tulane University Medical School, class of 2020.
66 Chen, D.W. (2018). Surprise Gift: Free Tuition for All N.Y.U. Medical Schools. The New York Times.
https://www.nytimes.com/2018/08/16/nyregion/nyu-free-tuition-medical-school.html
17
subsidies Medicare pays to teaching hospitals to train residents.67 If the IME adjustment to inpatients
DRG payments were halved, as MedPAC has recommended, over $3 billion would be made available
annually.68 To improve the availability of primary care services some of these funds could be deployed
in a direct subsidy program of the type outlined above, which we believe would be more effective than
the supply-side reforms recommended by the IOM. If a reduction in Medicare GME payments were to
be the source of funding this program, there should be a requirement that a PCP who receives the loan-
forgiveness subsidy should see a minimum number of Medicare and Medicaid patients. Including
Medicaid patients in the minimum requirement would have the advantage of encouraging
pediatricians to participate in the program.
A loan-forgiveness program would be a useful complement to reforms of the Medicare RBRVS. The
subsidy program would work more quickly and students could have more confidence in its long-term
impact for them. The specialty mix problem is enormous and our track record in reforming the RBRVS
(it is done infrequently and the gains have been eroded over time) is spotty, which means that while
RBRVS reform is the ultimate policy to address the issue, we should not hesitate to pursue this
additional tool.
67 Edmunds M., Sloan F.A., and Steinwald, A.B. (2012). Geographic Adjustment in Medicare Payment: Phase II: Implications for
Access, Quality, and Efficiency. Institute of Medicine of the National Academies, National Academy Press, 2012.
68 In fairness to MedPAC its recommendations have been more oriented to redeploying GME payments in hospitals to be more
accountable for achieving policy objectives rather than removing the subsidy dollars from hospitals. Nevertheless, redirecting some
of these dollars to reducing the income gap would, in a large sense, retain the money within the medical education system. See
MedPAC’s June 2010 for a summary of its recommendations for reprogramming GME payments.
18
Conclusion
The principal findings of the preceding analysis may be summarized as follows:
The mix of physicians in the US has too few PCPs and too many specialists.
The income gap between PCPs and specialists is a major determinant of the PCP/specialty mix.
Medicare physician payment policy is a major contributor to the income gap.
While Medicare GME payment policies do not appear to be a potent force in specialty
determination, they also do not provide a counterforce to the influence of the gap.
Changes in GME payment to hospitals to favor training of primary care physicians have little
potential to make a meaningful difference because other incentives affecting physicians-in-
training and teaching hospitals are too powerful. Redirecting these resources towards loan
forgiveness for those medical students who pursue careers in primary care appears more
promising.
But none of the approaches focused on medical education are likely to be fully successful unless
the Medicare relative value scale is also revamped so that it contributes less to the income
gap.69
We acknowledge that the income gap and specialty imbalance is part of a larger need to examine the
health care workforce in the US and related policy issues such as access to care in underserved areas.70
Nevertheless, the influence of Medicare on the US health care system as a whole provides an
opportunity to use Medicare payment policy to achieve broad social objectives such as influencing the
specialty choices made by doctors in training when they decide in what field to practice medicine.
69 In addition to the analysis above, our informal interviews of hospital executives and health policy experts support our view that
the incentives embodied in the Medicare fee schedule tend to overwhelm all other incentives, such as government funded workforce
programs and Medicare subsidies to teaching hospitals.
70 The Affordable Care Act of 2010 required the Comptroller General to appoint a 15 member National Health Care Workforce
Commission but, because the Commission was never funded, it has never met.
The USC-Brookings Schaeffer Initiative for Health Policy is a partnership between the Center for Health Policy at Brookings and the USC Schaeffer Center for Health Policy & Economics and aims to inform the national health care debate with rigorous, evidence-based analysis leading to practical recommendations using the collaborative strengths of USC and Brookings.
Questions about the research? Email communications@brookings.edu.
Be sure to include the title of this paper in your inquiry.
© 2018 The Brookings Institution | 1775 Massachusetts Ave., NW, Washington, DC 20036 | 202.797.6000